We often selectively interpret evidence to fit with our prior beliefs

Supporters and opponents of capital punishment read the same balanced account. Afterwards, both groups held stronger versions of their original views.

Participants accepted evidence that supported their position and scrutinised evidence that challenged it. The same asymmetry appears when a successful segment proves an experiment while a failed segment is dismissed as unusual, or praise in a survey is treated as candid while criticism becomes “context-dependent”. Prior beliefs help people organise information. The problem arises when the standard of proof changes with the direction of the result. Friendly evidence needs one plausible route; hostile evidence must close every escape.

A balanced document therefore cannot guarantee a balanced conclusion. It may give each side more material for defending what it already believed.

Evaluation rules are most useful before the outcome is known. Which result would weaken the favoured explanation? Which source would remain credible if it supported the opposing case? What evidence would end the project instead of producing another request for time? One person can also be asked to construct the strongest unwanted interpretation, using terms an intelligent opponent would recognise. That work is uncomfortable because it prevents the preferred view from choosing both the evidence and the rules under which the evidence will be judged.